Professional Documents
Culture Documents
Translation Memory - A Teaching Challenge
Translation Memory - A Teaching Challenge
Perspectives
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t794297831
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
205
eventually put on the market and have enjoyed growing popularity. Many com-
panies and virtually all international organisations as well as most professional
translation agencies handling large quantities of translation use TM systems as
aids.
The definition
TM stands for ‘Translation Memory’. The basic element in TM is the automa-
tion of the reuse and recycling of previously translated terms and sentences.
A TM database system automatically stores translations and allows for their
reuse. The database created is called ‘translation memory’. The translation unit
saved in the translation memory consists of a source language segment and its
target language equivalent.
TM overlaps with machine-assisted human translation and computer-assist-
ed translation in that they all depend on computers to improve or speed up the
translation process. (Hedden 2002)
In 1966, the Automatic Language Processing Advisory Commi�ee reported
that fully automatic, high-quality machine translation was impossible, and in-
stead, the commi�ee “recommended the development of computer-based aids
for translators”. (Somers 2003b:14)
In the late 1970s, the first proposal was made for “what is now called trans-
lation memory, in which previous translations are stored in the computer and
retrieved as a function of their similarity to the current text being translated”, a
tool then known as “Repetitions Processing”. (Somers 2003b: 14) The 1980s saw
the development of computational linguistic techniques and Alan Melby “pro-
posed the integration of various tools into a translator’s workstation”, such as
word-processing, terminology management, and automatic dictionary lookup.
(Somers 2003b: 14)
In 1980, Martin Kay suggested that translators might “examine past and fu-
ture fragments of text that contain similar material”, (1980: 19) and thus estab-
lished the basic notion of TM in the first of a small number of studies on the
topic in the 1980s. However, “it was not until the mid 1990s that TM systems
became commercially available, and then in a short period of time they were
quickly accepted.” (Somers 2003c: 33) Consequently, the mid-1990s saw the re-
lease of numerous TM products, such as Trados, Translation Manager (IBM),
and Déjà Vu (Atril).
The TM concept
A TM system is based on the reuse or recycling of previously translated terms
and sentences. In a TM system, translators can “consult a database of previ-
ous translations, usually on a sentence-by-sentence basis, looking for anything
similar enough to the current sentence to be translated, and can then use the re-
trieved examples as a model”. (Somers, 2003c: 31) TM assists human translators
by searching databanks for similar or exact previous translations. In principle,
human translators thus need to translate these same words, phrases, or contex-
tual segments only once. For areas in which there are large databanks and for
which source texts and translations require regular updates or revisions, it is
Downloaded At: 23:43 18 February 2009
obvious that TM can reuse stored data and thus increase efficiency. As Bert Es-
slink suggests in reference to localisation, this helps “reduce time to market of
localized versions drastically”. (Esselink 2003: 80)
It is not surprising then that TMs have been introduced in many workplaces,
spanning from international organisations to small-sized professional transla-
tion agencies, since the 1990s. Once all of the 3,000 translators at the European
Union organisations could access shared databases, these have constituted com-
mon and shared pools of reference.
In order to exemplify the usefulness of such networking, we posit that
Translator A has completed the English-to-Chinese translation of the sentence
The above mentioned business should be managed in accordance with the terms and
regulations concerned and that this translation is immediately stored in the data-
bank. Translator B is translating the sentence The business should be managed fol-
lowing the terms and regulations concerned and translator C is preparing to trans-
late the sentence The above mentioned affairs should be managed in accordance with
the terms and regulations concerned. Instead of laboriously working out their own
renditions, translators B and C can retrieve translator A’s translation 上述業務
應依照相關法令與規定執行 from the databank. What is more, if the context of
their respective translations demand so, B and C can modify A’s translation and
these renditions will then also be stored in the databank.
TM databanks also ensure that, e.g., specialist terminology is consistent with
different translators. Such terminological consistency streamlines the final edit-
ing. Today, no large translation project completed by a team of translators can
be done without TM tools.
Terminology management
A TM system enables translators to create term bases, add or delete terms,
searches, and filters. During the interactive translation initiated by a TM system,
users can search local or remote term bases and then insert the required transla-
tion of terms in the document they are translating. In a multi-user environment,
the term base server system allows users to connect to remote or on-line term
bases.
ses offer translators, agencies, as well as project managers some reference for
estimates of costs, workload, and translation deadlines.
The use of a TM system can save 30% to 60% of the total time and cost of
translation, depending on the type of text and the memory database. Lynn E.
Webb (1998-1999) summarises the savings of companies, translation agencies,
and companies with in-house translation departments a�er they employed TM
systems, as follows:
Table 1
Translation agency savings and profits a�er using TM tools
Without TM Using TM Savings
Number of words in project 40,000 40,000 N/A
Turnaround, in days 4-5 3 1-2
Total translation cost (Per project) USD 6,112 to USD 2,609 USD 3,503 to
USD 7,824 USD 5,215
Cost per word USD 0.15 to USD 0.07 USD 0.08 to
USD 0.19 USD 0.12
Total profit USD 576 to USD 5,936 to
USD 2,312 USD 7,556
Downloaded At: 23:43 18 February 2009
The Table indicates that the savings in terms of time and money for each
translation project ranges from USD 3,360 to USD 5,215, and that the savings on
individual words ranges from USD 0.08 to USD 0.12, at one of the companies
studied by Webb (1998-1999).
Acceleration of localisation
TM systems can be used for localisation when a new product with instruc-
tions wri�en in several foreign languages is released. TM systems, such as Tra-
dos, automatically convert dates, numbers, and measurements based on the
user se�ings across countries. Thus, TM use shortens the turnaround time of
localisation and the time to market.
translated sentence is automatically added and saved. When the same (or the
next) translator works on a text that has been translated before, the la�er can
be loaded into the database and the newly translated texts are then added. This
method is open to any user. The following example is illustrated with a (low)
65% match:
Figure 1
Trados window showing 65% match with previously translated sentences
In the figure, the Trados window shows the partial match of the sentence to
be translated compared with the previously translated sentence.
Downloaded At: 23:43 18 February 2009
The second method exploits the alignment function, which makes it possible
to match source-text and target-text segments that are “sentences or other more
or less easily distinguishable text portions, such as titles” (Somers 2003c: 34)
into bilingual pairs, segment by segment. Some language pairs, e.g., English
and Chinese, are so different in their syntactical structures and grammar that
it is very difficult to align them accurately. However, some TM systems allow
users to redo or modify incorrect alignments.
The third method is to import compiled translation memory databanks into
the trainee’s workstation. Despite the fact that there are many different TM
products, most use a common interchange format, which means that “TM data-
bases using one product can be imported into another.” (Somers 2003c: 34)
(a) The source text: “The army, driven back into Virginia, went into winter
quarters on the Rapidan - a tired, depleted army since the defeat at Get-
tysburg - and as the Christmas season approached, Ashley came home
210 2005. Perspectives: Studies in Translatology. Volume 13: 3
on furlough.”
(b) 聯盟軍既被打回佛金泥,便都歸到剌匹丹河上的冬令營來了。他們自從
葛的斯堡吃了那麼一個大敗仗,元氣業已大虧,並且疲倦得不能不休息
了。因此將近耶誔節的時候,希禮便也請假回來。
Created on: 2005/9/14, 00:14 Created by: ALGN Changed on: Changed
by:
(c) 自從葛底斯堡一戰吃了敗仗以後,部隊元氣大傷,師老兵疲。大軍被迫
退到佛吉尼亞,就在拉皮丹河畔安營過冬。快到耶誔節時,阿希禮回家
來休假了。
Created on: 2005/9/14, 01:12 Created by: ALGN Changed on: Changed
by:
(d) 南軍被趕回佛吉尼亞後,駐紮在拉皮丹河上的冬營裡──自葛底斯堡大
敗之後南軍就一蹶不振,元氣大傷。耶誔節轉眼又至,阿什禮回來度假
了。
Created on: 2005/9/14, 01:14 Created by: ALGN Changed on: Changed
by:
Downloaded At: 23:43 18 February 2009
When students compared these translations, they found that the translator of
(d) had the most concise style, because he/she translated ‘defeat’ as ‘大敗’ (two
Chinese characters) instead of (c)’s four-character ‘吃了敗仗,’ and (b)’s nine-
character ‘吃了那麼一個大敗仗’. On the other hand, some Chinese readers may
favour the second translation ‘吃了敗仗’ (c), because its style is more vivid.
Table 2
The concordance output of the use of 請/please in English and Chinese in
guidelines
1 疏散後,請協助其他乘客沿著緊急升降梯下降。
A�er evacuating,____ assist other passengers coming down the emergency slide.
Created on: 2005/9/18, 15:29 Created by: ALIGN! Changed on:
2 請協助其他旅客直接下機。
___ direct other passengers to evacuate away from the aircra� immediately.
Created on: 2005/9/18, 15:29 Created by: ALIGN! Changed on:
3 請將行李放置在頭頂上的行李放置艙內或是座位下方。
Please stow the baggage you bring with you in the overhead bins or under the
seat in front of you.
Created on: 2005/9/18, 15:29 Created by: ALIGN! Changed on:
4 為確保閣下與其他旅客的飛行安全與順利,請與我們配合,謹慎遵守基本法則和
指引。
To ensure a safe and pleasant journey for you and your fellow passengers, we ask
for your cooperation in carefully observing some basic regulations and guide-
lines.
Created on: 2005/9/18, 15:29 Created by: ALIGN! Changed on:
Downloaded At: 23:43 18 February 2009
Term-base buildup
Teachers can also use the TM system to teach students to use the terminology
management systems to develop term bases in different subjects. Term bases
mainly save words and phrases, whereas databanks save texts. Sophisticated
term bases allow translators to design and store a variety of entries, including
a lexical item’s grammatical features and sample sentences, and extralinguistic
information about a lexical item, such as definitions and background data. In
addition, trainee translators may learn the principles behind such systems for
compiling term bases for customers regarding the same subject. This may also
be used for demonstrating to students how different companies use the same
specialist terms in different ways.
Cost-effectiveness assessment
Furthermore, TM systems can be used to show prospective translators how
to calculate the cost of their work and how much they should charge for texts in
which there is much repetition. Statistical techniques supported by the TM sys-
tem can analyse the similarity of new source texts to the corpus already stored.
Thus, TM systems allow users to establish how many translated segments cor-
respond to the source-text segments. This, in turn, permits translators to assess
immediately the amount of repetition of terms in the documents. This is obvi-
ously much easier in translations between Indo-European languages, such as
English and French, for which words and sentence structures are more alike.
212 2005. Perspectives: Studies in Translatology. Volume 13: 3
Problems with TM
In the above, I have outlined only a few methods that can be used in order
to familiarise and train translator trainees in TMs. However, this should also be
balanced by pointing out the weaknesses of the systems.
Whereas a TM databank is useful for recycling and thus lightens the work-
load of translators, the retrieval does not distinguish between adequate and
erroneous translations that are saved in a TM databank. This means that the
individual user will have to critically review any previously translated segment
that is displayed for approval or rejection – and may have to reject otherwise
Downloaded At: 23:43 18 February 2009
adequate renditions that are merely inadequate in the new context. It should be
stressed to students that, in principle if not in practice, one method of improv-
ing the results stored in databanks is to have all data reviewed for accuracy and
appropriateness before it is stored. From a long-term perspective, revisers might
check the translations stored in the databanks, in some kind of cooperation with
engineers. One possible short cut is to align source-language and target-lan-
guage texts and use the aligned segments for developing a TM databank. But
this should only be done provided the target-language segments are checked
for adequacy and accuracy. However, at the present stage of development, I do
not believe we can get any further than this.
Advantages of TM training
Learning TM strengthens students’ confidence about their future careers in
the market place, with translation agencies, international globalised compa-
nies, and localisation industries. The pedagogical advantages of TM training at
universities are summarised in Table 3 and are adapted from Kathleen Graves
(2003: 83).
Chung-ling. Translation Memory: A Teaching Challenge. 213
Table 3
The pedagogical advantages of TM use.
Areas Pedagogical Advantages
Knowledge 1. The important financial advantages of TM tools.
2. The lexical, syntactic, and stylistic differences between different text-
types, the individual choices made by translators, and appropriate
translation models by means of TM-driven Concordance.
3. Real-time data recycling and shared information.
Awareness 1. The use of TM tools to translate technical texts in an efficient and cost-
effective way.
2. The use of TM in translation.
3. The ability of TM tools to manage the budget, the process, and to assure
quality and terminological consistency in translation projects.
Skills 1. How to establish TM databanks and customer-tailored term bases.
2. How to use TM tools to translate web pages.
3. How to use exact or fuzzy matching functions in TM systems to access
previously translated sentences.
4. How to use TM-driven concordances.
5. How to use fuzzy or exact searches to retrieve terms from term bases.
6. How to use the alignment tool to compile parallel bilingual corpora.
Downloaded At: 23:43 18 February 2009
Conclusion
TM systems are used extensively in professional translation and translator
training institutions have to introduce students to them. Disregarding the prob-
lem that in every country TM will meet with specific problems in terms of the
size of databanks available and specific translation problems depending on the
language pairs involved, teaching staffs, at present, are hampered by the small
size of the corpora available for training purposes.
It would be a great advantage if specialised and professional TM databanks,
e.g., those used by globalised companies for the translations of their equipment
and manuals, were accessible – or at least cheap - to training institutions. This
would be a great advantage for all would-be translators, as they could use such
material to become familiar with the challenges and work procedures in the dai-
ly lives of professional translators in today’s world. One could imagine that this
might also be beneficial to those who offered such access to translation schools,
214 2005. Perspectives: Studies in Translatology. Volume 13: 3
since this would minimise the in-house training that they are otherwise obliged
to give to newcomers.
No ma�er whether such chances of hands-on experience are offered or not,
translation training at universities must keep up with the rapid changes in the
industry. The integration of TM components into current translation educa-
tion is imperative; the “resistance to change can represent a major impasse”.
(O’Brien 1998: 121) TM training not only benefits future translators, but also
enables industry to recruit qualified translators.
Works cited
Abaitua, Joseba. 1999. Is it worth learning translation technology? = www1.
Aula.int. 2005. Translator Training and Modern Market Demands. Perspectives: Studies in
Translatology 13. 132-142.
Esselink, Bert. 2003. Localization and translation. In: Somers, Harold (ed.). 67-86.
Graves, Kathleen. 2000. Designing Language Courses. London: Heinle & Heinle.
Hedden, Thomas D.1992-2000. Machine translation: A brief introduction. = www2.
Heyn, Ma�hias. 1998. Translation memories: Insights and prospects. In: Bowker, Lynne
et al. (eds.). 123-136.
Kay, Martin.1980. The proper place of men and machines in language translation. Re-
search Report CSL-80-11, Xerox PARC, Palo Alto, Calif. Reprinted in: Machine Transla-
tion 12. (1997). 3-23.
O’Brien, Sharon.1998. Practical experience of computer-aided translation tools in the
Downloaded At: 23:43 18 February 2009
Websites
www1 = h�p://sivio.deusto.es/abaitua/konzeptu/ta/vic.htm (accessed 11 March 2002).
www2 = h�p://www.he.net/~hedden/intro_mt.html (accessed 11 March 2002).